基于超像素的面向对象遥感图像分类方法分析-analysis of object-oriented remote sensing image classification method based on superpixel.docxVIP

AbstractWiththedevelopmentofremotesensingimagingtechnologyandtheimprovementofmulti-sourceimagesresolutioninsatellitevisiblelight,multi-spectralandhyperspectral,thehighresolutionremotesensingimagehasbeenwidelyusedinvariousfields.Thetraditionalsceneclassificationmethodbasedonpixelislackofaccuracytomeettherequirementofhighresolutionremotesensingimageprocessing,becauseofitsinabilityofextractingthesameobjecttypesstructureandtexturefeatures.Sotheobject-orientedremotesensingimageclassificationmethodisproposedtosolvetheaccuracyproblem.Object-orientedremotesensingimageclassificationincludeobjectsegmentationandobjectclassification.Itpresentsthehighresolutionremotesensingimagesegmentationtechnology,anddiscussesthesuperpixelsegmentationalgorithmbasedonSLIC.Thisalgorithmproducesobjectswithstronghomogeneityandcompactstructure,andextractsfeaturesoftheseobjectsthatclearlyreflectthetypicalstructuralinformationandspatialtopologicalrelationship.Thusthealgorithmprovidesefficientandreliablesegmentationstrategy.Thispaperalsopresentsanobject-orientedremotesensingimagesceneclassificationmethodbasedonsuperpixelsegmentation.The methodisconsistoftypicalobjectsclassificationgeneration,superpixelsegmentationbasedonSLIC,textureandspectralfeaturesextractionoftypicalobjectsandsupervisingclassificationbasedonsupportingvectormachine.Remotesensingimageclassificationsoftwaresystemisdesignedandimplementedtomeettherequirements.Thispaperdiscussedtheinfluenceoffeatureextraction, clusterand thescale ofsegmentation.Theclassificationexperimentsonsomevisiblelightand infrared imageshowthatthemethodhasgoodapplicabilityinhighresolutionremotesensingimagesceneclassification.Keywords:HighResolutionRemoteSensingImage,Object-OrientedClassification,SuperPixel,SimpleLinearIterativeClustering,SupportingVectorMachine目录摘要...............................................................................................................IAbstract...................................................................................